1. Introduction

Six years after the first report of a perovskite solar cell, the power conversion efficiency of small devices has increased from merely 3% ^[1] to over 22% ^[2]. This is the most impressive advancement in the history of photovoltaic technologies. In comparison, it took 40 years for crystalline silicon research cells to get over 22%, starting from 5%. However, the reported record efficiencies for perovskite solar cells generally come from devices with a very small active area of around 0.1 cm^{2 [3-6]}. Depositing flat, uniform and fully covered perovskite thin-film by industrial available approaches is still a thorny issue.

Previously, we reported a facile and scalable gas pump method by accelerating the solvent evaporation to speed up the precipitation of perovskite, and deposit extremely dense and uniform perovskite thin-films. We prepared planar perovskite solar cells by using this process, reaching the efficiency up to 19% with an average efficiency of (17.38±0.7)% ^[7]. The perovskite films were fabricated in air conditions with a relative humidity of 45%-55%, which would have a prospect in industrial applications of large-area perovskite solar panels. We also applied this method on PET-ITO flexible substrates and fabricated flexible perovskite solar cells with efficiencies up to 11.34%^[8].

This work is about our up-scaling efforts on this gas pumping method. By optimizing the equipment and techniques, we are able to fabricate large area perovskite modules of 17.3 cm² active area with over 10% power conversion efficiencies.

2. Device fabrication

The architecture of the module is shown in Fig. 1. A compact zinc oxide (ZnO) layer of about 50 nm thick was coated on top of the fluorine doped tin oxide (FTO) transparent electrode by magnetron controlled plasma sputtering. The compact ZnO layer acts as the collector of electrons and the blocking layer for positive charges. Schematic procedures to prepare the perovskite films by the gas pump method are shown in Fig. 2. First, 40% CH₃NH₃PbI₃ precursor solution in N, N-dimethylformamide (DMF) was casted with a slotdie coater on top of the ZnO layer. Then the substrate was put into a sample chamber connecting to the gas pump system. The system is comprised of a large low pressure chamber and a sample chamber, these chambers are connected by gas drainage pipes controlled by valves. By opening the valves connecting the sample chamber and the low pressure system maintaining at 100 Pa, the fast pumping of the sample chamber leads to rapid evaporation of the DMF solvent. Brown perovskite films can be obtained with 20 s. The color of the film became dark brown after being annealed at 100 ℃ for 10 min.

After annealing, a carbon layer was screen-printed on top of the perovskite film from a home-made carbon paste. This carbon layer acts as a collector for positive charges and back contact. These layers were scribed by laser into eight strips, forming seven cells with series connection (Fig. 1).

3. Performance and discussion

Photocurrent-voltage (J-V) characteristics of the devices were measured by employing a Keithley 2400 source-measure under illumination of 100 mW/cm² by a 450 W class AAA solar simulator equipped with an AM1.5G filter (Sol2A, Oriel Instruments). The exact light intensity was determined by a standard silicon reference cell (91150V, Oriel Instruments). A metal mask which was a little smaller than the nominal active area of the device defined as the overlap between FTO and the carbon electrode was applied when measuring the solar cell efficiency, as to avoided edge effects.

The J-V curve of the best performing $5 \times5$ cm² module (Fig. 3(b)) exhibits the short-circuit current density ($J_{\mathrm{SC}})$ of 3.25 mA/cm², the open-circuit voltage ($V_{\mathrm{OC}})$ of 6.14 V, the fill factor of 0.53 and the power conversion efficiency (PCE) of 10.6%. Since the modules consisted of seven cells connected in series, the module voltage is the sum of these cells, while the module current is limited by the lowest one of them. Therefore, the homogeneities of all the four layers (FTO, ZnO, perovskite and carbon) became decisive for the performance of the module. As shown in Fig. 3(c), the average PCE of 50 modules is $(9.97 \pm0.35)$%, showing that our film processing techniques are well controlled and reproducible.

We may also expect the resistance of the series contacts to play an important role in the solar-electricity generation. If the contact resistances become significant, the series resistance of the module will rise and hence reduce the fill factor. To verify this effect, single cells (1 cm² active area) of the same architecture were fabricated for comparison. As shown in Fig. 3(a) and Table 1, the PCE of the best single cells reached 15.1%, which is significantly higher than that of the modules. By comparing the data given in Table 1, we found that the current outputs for the small device and module are similar (3.25 mA/cm² × 7 = 22.75 mA/cm²), while the V_OC and fill factor of the small device are larger (1.05 V×7 = 7.35 V).

Looking into the SEM image taken on the cross section of the device (Fig. 4), we can see that the carbon layer consists of large graphite flakes of several micro-meters in size filled with a lot of carbon black particles. The main conducting pathways are formed by graphite flakes which make contacts to each other. Numbers of hollows can be found between the carbon layer and the perovskite layer, which will obviously hinder the charge collection. Densely distributed hollows are also found between the contact of carbon and FTO, which should result in a high contact resistivity, adding series resistance into the module, leading to a lowered fill factor and $V_{\mathrm{OC}}$.

The operating lifetime is also a major concern for perovskite solar cells. We sealed the module with plane glass and water-proof sealant. The sealed devices were placed outdoors and tested with a certain time interval. The well capped devices were rather stable through the changing weather of the tropics in south China. As we recorded in Fig. 5, for over 140 days we did not see any trend of degradation. The fluctuation of PCE is the result from the changing solar intensity in different weather conditions.

Based on the techniques applied on the 5×5 cm² modules, we further increase the module size to 45×65 cm². A demonstration power station made of 32 perovskite panels was set up on our site (Fig. 6). These works on module fabrication and characterization are still going on, and will be published elsewhere.

4. Conclusion

We have developed a gas-pumping method for fabricating unique and compact perovskite film in large area. Based on these techniques, we are able to produce large area perovskite modules from $5 \times5$ cm² to $45 \times65$ cm². The power conversion efficiencies for $5 \times5$ cm² modules reached 10.6% with good reproducibility. No significant degradation was found after 140 days of outdoors testing. The great drop of PCE from single cells to modules can be attributed to large series resistivity resulting from the poor contact between carbon and TCO.